AI's next winners may be the adopters, not the suppliers

The $730 billion starting point

The framing starts with scale. In an October 15 note, BNY Mellon's Walter Scott arm, via Jay McLeod, puts the AI capex wave at roughly US$730 billion from leading technology companies in 2026, much of it aimed at AI infrastructure — "one of the largest investment cycles in decades." McLeod's caution is historical: "the ultimate beneficiaries of transformative technologies are not always the businesses supplying the tools." The corollary: as the technology matures, value creation broadens beyond the tool suppliers, and the durable edge belongs to companies pairing AI with advantages that are hard to replicate — proprietary data, deep operational expertise, strong customer relationships, established market positions. m:…8150

The note sketches where that adoption lands: beyond the data center, intelligence gets embedded into factories, vehicles, and industrial equipment, with applications emerging across financial software, energy, utilities, manufacturing, and transportation. Rather than disrupting incumbents, McLeod argues, AI may reinforce their competitive positions by improving decision-making and customer outcomes. The implication, in the note's logic, is that the next leg of value belongs to operators, not just builders — a point aimed squarely at portfolios overweight the picks-and-shovels trade. m:…8150

Picks and shovels: power and cooling

Wellington Management's Andrew Heiskell and Brian Barbetta, writing October 16, offer a parallel map: think of AI as a connected ecosystem of seven categories of companies — hyperscalers, infrastructure suppliers, productivity beneficiaries among them — rather than a binary list of winners and losers. Their emphasis lands on the literal picks and shovels: power generators, turbines, electrical equipment, cooling systems, construction and excavation equipment "required for the massive expansion of computing power." With capital spending sustained, they argue the investment case for that layer remains strong. m:…1822

Two caveats come with the map. First, hyperscaler sentiment has followed a push-forward-then-fall-back pattern over the past two years, so volatility is the norm, not the exception. Second, industrial AI stocks are not a single trade — there have been relative winners and losers over the last 12 months, and the authors expect dispersion to increase. They also contend the market is underestimating the scale of the businesses being built and the profit growth that will follow. m:…1822

The adopter case, with a name on it

Baillie Gifford gives the adopter thesis a concrete example: Salesforce's Agentforce. The firm's Global Alpha strategy added Salesforce to its portfolio, citing three edges — distribution, data, and engineering. More than 150,000 organizations already use Salesforce products, from Disney to the Danish government; many hold significant volumes of data inside Salesforce systems, giving Agentforce secure access to high-quality, well-structured records; and the company's scale helps it attract top engineering talent plus an external developer ecosystem. Agentforce itself launched in late 2025, using generative AI to understand natural language and resolve customer service issues, handling multiple interactions at once and operating around the clock. m:…5561

The pattern across the three notes is convergence, and that reading is the author's synthesis, not any single firm's claim: BNY Mellon argues the beneficiaries shift from suppliers to adopters, Wellington maps the ecosystem and keeps a sleeve for the power-and-cooling layer underneath it, and Baillie Gifford names an adopter with distribution and data. None of the three is calling the infrastructure trade over — the $730 billion figure is their shared starting point. The shared move is a broadening of where professionals should look next: toward the companies that turn the buildout into productivity, and the industrial layer that keeps it running. That is an allocation argument, not a recommendation — and the notes agree leadership within the ecosystem will keep shifting. m:…8150 m:…1822 m:…5561